Bugfixes for coh_tmm and gradient calculation

Small bugfixes in

• vectorized_tmm_dispersive_multistack
• check_inputs
• is_not_forward_angle

directly import vectorized_tmm_dispersive_multistack from tmm_fast as coh_tmm

tmm_fast v0.2

First pre-release for the tmm_fast package to make it available for PyPi

what is provided:

coherent vectorized transfer matrix computations which are differentiable via autograd
a gymnasium environment that interprets creating a thin-film as an MDP
PyTorch GPU acceleration for transfer matrix computations
Thin-film dataset generation with Dask
Planned features:

incoherent / coherent transfer matrix computations
testing and CI/CD integration